On MetaHire, my team graduation project, I designed and trained a CNN-LSTM vocal-tone model using MFCC features from the RAVDESS dataset. I also implemented inference and evaluation for DeepFace emotion analysis and L2CS-Net gaze estimation, combining the three Python modules through a weighted fusion rule.
For my Australian Rain Prediction project, I analyzed weather records, applied PCA, and compared classification models using cross-validation. Random Forest produced the strongest evaluated result, with 84.7% accuracy.
My training at Digital Egypt Pioneers Initiative (DEPI) and National Telecommunication Institute (NTI) gave me practical experience in machine learning, computer vision, and data analysis. I also fine-tuned YOLOv5 for tennis-ball detection and used pretrained YOLOv8x for player detection and tracking.
